What Does 'High Availability' Mean in Live Streaming?
Learn more about SLAs at the next Content Delivery Summit.
Read the complete transcript of this clip:
Steve Strong: What does, what does "high-availability" even mean? Everyone talks about SLAs and people talk about five nines and such like. We thought it'd be interesting just throw some numbers up to get people to think a little bit about what these numbers mean. Cause it's easy to say 99s SLA, but what, what the heck does that mean? It means you're up for 99.999% of the time. But what does that even mean? Is that over the course of a year, or over the course of a day, or over the course of a 100-meter final? The answer is very different. So you need to know how you're being measured and you need to know what the implications are. If you've got a very high SLA and you're doing a sporting event and it's a wet afternoon with nothing happening and you go down, no one's probably going to care, even though technically you breached your SLA. If you go down in the middle of the 100-meters final, people are going to care a lot, even if you don't breach your SLA.
So if you look at the numbers--90% per year, that's easy. You can be down for 36 days--a walk in the park. So no one here is interested in that. At the other end of the scale, if you're at 99s and you're being measured on a daily basis, you can be down for 0.86 seconds a day and that's not very long to handle any sort of error. You really don't have much time to respond there. Relating that to the 100-meters final, if you're up for 99.999 in the final, you don't lose a millimeter. That would be okay.
But the reality is, that isn't going to be what happens. Let say it's your daily outage, and let's say you can get back in time, 0.864 seconds. You still miss the guy crossing the line. That's a shame.
So you really need to know what these numbers mean. You need to know how they apply to your customers and what your customers mean when they ask for these numbers.
Adrian Roe: You also need to kind of think a little bit about how it applies to your suppliers. So, for example, it's very common with traditional systems just to back to back your SLAs. I've got a big old enterprise SLA and I'm gonna talk to providers are going to give me enterprise SLAs, and someone like AWS, they've upped their SLA in the last couple of years because it changed since the last time I looked. And I'll give you a four-nines SLA now, which is great. Now, they're actually very explicit about how they measure that. They measure that over the course of a month, which means that they can be down for 4.3 minutes. And then if they go a second over that they will give you EC2 credits to the tune of a second. So if they're down for your 100-Meter Final, they'll go, "We're terribly sorry about that. Here's 12 cents." And that's probably not very useful. So it used to be when we were building banking systems, you'd say, "Here's the big SLA, and here's another big SLA, and here's another big SLA. And then you kind of chain them all together and you say, "Oh. Hence, the big SLA just doesn't hold anymore."
id3as Directors Steve Strong and Adrian Roe discuss what's required to maintain 99.99 and 99.999 uptimes in this clip from their presentation at Content Delivery Summit.